This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Will tree-ssa be GCC 3.5?
- From: Joe Buck <jbuck at synopsys dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: steven at gcc dot gnu dot org, gcc at gcc dot gnu dot org
- Date: Mon, 1 Dec 2003 09:36:51 -0800
- Subject: Re: Will tree-ssa be GCC 3.5?
- References: <10312011515.AA26277@vlsi1.ultra.nyu.edu>
On Mon, Dec 01, 2003 at 10:15:12AM -0500, Richard Kenner wrote:
> The question of to what extent optimizations will be improved by doing
> them at a higher level is one that can only be answered by doing them.
> Likewise for the cost of such optimizations.
Ten years of experience suggests that certain optimizations needed for
decent C++ performance will never be done in the current GCC framework.
tree-ssa is pretty much the only hope.
And C++ is only one example, any language with record types will have
similar issues.
> There has been a lot of speculation, but nobody can say with certainty
> whether or to what extent any of the current RTL optimizers can be removed
> without losing significant code performance, for another example.
That's a separate question and I agree that the answer is not known.
However, tree-ssa does not require the removal of the RTL optimizers.